Fitting the multi-wavelength spectral energy distributions (SEDs) of galaxies is a widely used technique to extract information about the physical properties of galaxies. However, a major difficulty lies in the numerous uncertainties regarding almost all ingredients of the SED modeling of galaxies. The Bayesian methods provide a consistent conceptual basis for dealing with the problem of inference with many uncertainties. While the Bayesian parameter estimation method have become quite popular in the field of SED fitting of galaxies, the Bayesian model comparison method, which is based on the same Bayes’ rule, is still not widely used in this field. With the application of Bayesian model comparison method in a series of papers, we show that the results obtained with Bayesian model comparison are understandable in the context of stellar/galaxy physics. These results indicate that Bayesian model comparison is a reliable and very powerful method for the SED fitting of galaxies.

This paper summarizes my thoughts, given in an invited review at the IAU symposium 341 “Challenges in Panchromatical Galaxy Modelling with Next Generation Facilities”, about how machine learning methods can help us solve some of the big data problems associated with current and upcoming large galaxy surveys.

Modelling and interpreting the SEDs of galaxies has become one of the key tools at the disposal of extragalactic astronomers. Ideally, we could hope that, through a detailed study of its SED, we can infer the correct physical properties and the evolutionary history of a galaxy. In the past decade, panchromatic SED fitting, i.e. modelling the SED over the entire UV–submm wavelength regime, has seen an enormous advance. Several advanced new codes have been developed, nearly all based on Bayesian inference modelling. In this review, we briefly touch upon the different ingredients necessary for panchromatic SED modelling, and discuss the methodology and some important aspects of Bayesian SED modelling. The current uncertainties and limitations of panchromatic SED modelling are discussed, and we explore some avenues how the models and techniques can potentially be improved in the near future.

We show our work on the L-Galaxies semi-analytic models of galaxy formation, which includes the radial resolved distribution of star, gas, SFR and metallicity on each galaxy disk. The newest version of the codes include the H22-to-HI gas transition prescriptions and the chemical enrichment of various elements. Our revised model can give results on cold gas components, radial metallicity gradients and scaling relations, which can fit the recent observations.

At high redshift, the contribution of strong emission lines to the broadband photometry can cause large uncertainties when estimating galaxy physical properties. To examine this effect, we investigate a sample of 54 LBGs at 3 < zspec < 3.8 with detected [OIII] line emissions. We use CIGALE to fit simultaneously the rest-frame UV-to-NIR SEDs of these galaxies and their emission line data. By comparing the results with and without emission line data, we show that spectroscopic data are necessary to constrain the nebular model. We examine the K-band excess, which is usually used to estimate the emissions of [OIII]+Hβ lines when there is no spectral data, and find that the difference between the estimation and observation can reach up to > 1 dex for some galaxies, showing the importance of obtaining spectroscopic measurements of these lines. We also estimate the equivalent width of the Hβ absorption and find it negligible compared to the Hβ emission.

We present the physical properties of Lyα emitters (LAEs) in a “DLA-concentrated regions” where there are 3 or more DLA within (50 Mpc)3 cubic box. We observed LAEs in a DLA-concentrated region at z = 2.3, the J1230+34 field, with Subaru/Suprime-Cam. In the 50 Mpc scale, we found no deferences in properties of LAEs such as Lyα luminosity function in the DLA-concentrated region compared to other fields at similar redshift. On the other hand, we found a ∼10 Mpc scale LAE overdensity around a strong DLA with NHI = 1021.08 cm−2.

Recent discoveries of high-redshift galaxies have revealed the diversity of their physical properties, from normal star-forming galaxies to starburst galaxies. To understand the properties of these observed galaxies, it is crucial to understand the star formation (SF) history, and the radiation properties associated with the SF activity. Here we present the results of cosmological hydrodynamic simulations with zoom-in initial conditions, and show the formation of the first galaxies and their evolution towards observable galaxies at z = 6. In addition, we show their multi-wavelength radiative properties. We find that star formation occurs intermittently due to supernova (SN) feedback at z > 10, and their radiation properties rapidly change with time. We suggest that the first galaxies are bright at UV wavelengths just after the starburst phase, and become extended Lyman-alpha sources. We also show that massive galaxies cause dusty starburst and become bright at infrared wavelengths.

In this paper, I review several dust evolution studies based on the DustPedia nearby galaxy sample. I first present the dust spectral energy distribution model, implementing a hierarchical Bayesian method, that we have developed. I then discuss the dust evolution trends we have derived among (integrated) and within (resolved) galaxies. In particular, we show that the trend of dust-to-gas ratio with metallicity is clearly non-linear, indicating the need for grain growth in the interstellar medium. Our trend is closer to the one derived with damped Lyα systems than what was suggested by previous studies. We finally demonstrate the universal processing of small amorphous carbon grains by stellar photons.

In Panda et al.2018a, we constructed a refined sample from the original Shen et al.(2011) QSO catalog. Based on our hypothesis — the main driver of the Quasar Main Sequence is the maximum of the accretion disk temperature (TBBB) defined by the Big Blue Bump on the Spectral Energy Distribution (Panda et al.2017; Panda et al.2018b). We select the four extreme sources that have RFeII ⩾ 4.0 and use {CIGALE (Boquien et al.2018) to fit their multi—band photometric data. We also perform detailed spectral fitting including the Fe II pseudo—continuum (based on Śniegowska et al.2018)) to estimate and compare the value of RFEII. We show the dependence of FeII strength on changing metallicity.

Dust absorbs stellar emission and reradiates this energy in the far-infrared (FIR). FIR observations hence give us a direct view of the dust, and allow us to study its properties. Unfortunately, FIR observations are only available for a small subset of galaxies. In this work, we estimate the global FIR emission from global UV-NIR observations. We show that a machine learning method clearly outperforms a SED modelling approach. For each galaxy, we not only predict the FIR flux across the 6 Herschel bands, but also estimate individual uncertainties. We inspect the worst predictions, and investigate how the machine learning predictor generalizes on new data. Our predictor can be used as a virtual observatory, which is especially useful now that there is still no confirmed next-generation FIR telescope.

Recently huge amount of dust Mdust ≃ 106−7M⊙ in galaxies at z = 7–8 has been discovered by ALMA observations. The suggested timescale of the dust production was a few–several×108 yr, while the stellar mass was several × 109M⊙. This amount of dust cannot be easily explained only by a supply from supernovae if we consider the dust destruction by reverse shocks. We propose that these values can be consistently explained if we take into account the grain growth in the interstellar medium (ISM). This scenario successfully reproduces the evolution of the dust mass, as well as the SFR, and stellar mass simultaneously. We conclude that even at such an early epoch of the Universe, the dust grain growth in the ISM plays a significant role in galaxies.

Numerical simulations of disk galaxies with steady (long-lived) and dynamic (short-lived) spiral arms suggest that offsets between stellar and gas spiral arms depend on their nature or lifetime (Baba et al.2015). Based on this theoretical study, we investigated gas-star offsets in the nearby grand-design spiral galaxy M51, and found that its two spiral arms exhibit different offset dependences against radius. One arm is consistent with a steady arm, while the other is consistent with a dynamic arm. We deduce that this difference is likely due to a tidal interaction with the companion galaxy (Egusa et al.2017). For this study, a stellar mass distribution with a high accuracy at a high spatial resolution is essential, which has come to be available by applying recent SED fitting techniques to multi-wavelength images. We are now working to extend this study to other nearby spiral galaxies.

The spectral energy distribution (SED) model should treat the evolution of a galaxy from its birth. Dust in galaxies affects the formation and evolution of galaxies in various ways. For example, dust grains scatter and absorb stellar emitted ultraviolet (UV) photons and re-emit the radiation at infrared (IR) wavelengths. In this work, we construct a galaxy SED model based on our dust evolution model (Asano et al. 2013a,b, 2014) with a rigorous treatment of the chemical evolution. To reduce the computational cost, we adopt mega-grain approximation (MGA; (MGA; Inoue, 2005). MGA regards a high density dusty region as a huge size (10 pc) dust grain for calculating dust scattering. In this approximation, we can solve the radiative transfer easily and provide SEDs and attenuation curves of galaxies. This model can be used to fit any galaxy in the wavelength range of 10 nm-3 mm.

I will review the tips learned from panchromatic modeling of active galactic nuclei (AGNs), based on our recent work to study the relationship between AGN and star formation (SF). Several AGN SED models are compared, and a significant AGN contribution is found in the IR luminosities and corresponding star formation rates (SFRs). I will review the AGN-SF relation and how different parameters and sample selections affect the observed correlation. I will then report on the constant ratio discovered between the SFR and the black hole mass accretion rate (BHAR), and their implications on the gas supply and galaxy formation history of these systems. Caveats and important questions to answer are summarized at the end.

Recently, spectroscopic detections of O[III] 88 μm and Ly-α emission lines from the z ≍ 9.1 galaxy MACS1149-JD1 have been presented, and with these, some interesting properties of this galaxy were uncovered. One such property is that MACS1149-JD1 exhibits a significant Balmer break at around rest-frame 4000 Å, which may indicate that the galaxy has experienced large variations in star formation rate prior to z ∼ 9, with a rather long period of low star formation activity. While some simulations predict large variations in star formation activity in high-redshift galaxies, it is unclear whether the simulations can reproduce the kind of variations seen in MACS1149-JD1. Here, we utilize synthetic spectra of simulated galaxies from two simulation suites in order to study to what extent these can accurately reproduce the spectral features (specifically the Balmer break) observed in MACS1149-JD1. We show that while the simulations used in this study produce galaxies with varying star formation histories, galaxies such as MACS1149-JD1 would be very rare in the simulations. In principle, future observations with the James Webb Space Telescope may tell us if MACS1149-JD1 represents something rare, or if such galaxies are more common than predicted by current simulations.

Despite decreasing cosmic star formation rate density over the last 10 Gyr, the stellar mass (M*) buildups in galaxies were still progressing during this epoch. About 50% of the current M* density in the universe was built over the last ∼8.7 Gyr. In this research, we investigated the stellar mass buildup and quenching of spatially resolved regions within massive disk galaxies over the last 10 Gyr. We apply the spectral energy distribution (SED) fitting method to SEDs of sub-galactic regions in galaxies to derive the spatially resolved distributions of SFR and M* in the galaxies. This namely pixel-to-pixel SED fitting method is applied to massive disk galaxies at 0.01 < z < 0.02 and 0.8 < z < 1.8. We found that massive disk galaxies tend to build their M* and quench their star formation progressively from the central region to the outskirts, i.e. inside-out stellar mass buildup and quenching.

Stellar masses are crucial ingredients for putting galaxies in the context of galaxy evolution and are commonly evaluated via Spectral Energy Distribution (SED)-fitting analyses which are hampered by dust attenuation. Observational constraints of attenuation in various galaxy classes provide key inputs for fitting a SED. I will present recent results about the attenuation properties of a sample of Herschel-selected galaxies at 0.7 ⩽ z ⩽ 1.6 widely spanning the star-forming Main Sequence (MS). I will show that far-IR selected galaxies on the MS are well described with local attenuation recipes. Conversely, common recipes cannot recover the SFR of far-IR selected starburst galaxies well above the MS. The SFR of these outliers appears to be hidden by the ∼90% in optically thick cores. These findings pose challenges for SED-fitting codes based on energy balance assumptions that might break in these peculiar sources.

We have developed a new SED fitting tool specialized for frontier redshift galaxies. It is a common case for high-z galaxies that the available data are restricted to rich optical to near-infrared photometry and few far-infrared (FIR) data deep enough to detect the faint object (e.g., HST/WFC3 + Spitzer/IRAC + ALMA). In such situation, one cannot perform a complicated modeling of dust emission in FIR regime. We then adopt simple treatment for the dust emission using empirical LIRG templates. Instead, we adopt a sophisticated and physically motivated modeling for stellar and nebular emission parts in rest-frame UV-to-optical regime. Our new code fits not only broad band photometry but also spectral emission line flux. There is an option to fit observed SED with two templates with different physical properties. Our new code, PANHIT, is now in public, and was already applied to some high-z frontier galaxies.

NGC 300 is a near-optical twin of the Local Group galaxy M33, which are benchmarks for understanding late-type spiral galaxies. They are two bulgeless and low-mass spiral galaxies in different environments. In order to explore the common properties and differences between the two nearby low-mass systems, we first use the simple chemical evolution model to explore the star formation history (SFH) of NGC300 and M33, and then compare the feasible model predicted SFH of NGC 300 with that of M33. Through comparing the SFHs between them, it can be found that the mean stellar age of NGC 300 is older than that of M33, there is a recent lack of primordial gas infall onto the disk of NGC 300, recent star formation along the disk of NGC 300 is less active than that of M33, and the local environment may play a key role in the secular evolution of a galaxy.

Modern cosmological simulations suggest that the hierarchical assembly of dark matter halos provided the gravitational wells that allowed the primordial gases to form stars and galaxies inside them. The first galaxies comprised of the first systems of stars gravitationally bound in dark matter halos are naturally recognized as the building blocks of early Universe. To understand the formation of the first galaxies, we use an adaptive mesh refinement (AMR) cosmological code, Enzo to simulate the formation and evolution of the first galaxies. We first model an isolated galaxy by considering much microphysics such as star formation, stellar feedback, and primordial gas cooling. To examine the effect of Pop III stellar feedback to the first galaxy formation, we adjust the initial temperature, density distribution and metallicity distributions by assuming different IMFs of the first stars. Our results suggest that star formation in the first galaxies is sensitive to the initial conditions of Pop III supernovae and their remnants. Our study can help to correlate the populations of the first stars and supernovae to star formation inside these first galaxies which may be soon observed by the (James Webb Space Telescope JWST).